| |
The START Study Controversy: conducting a "When To Begin Study"
|
| |
| |
The START Study
Investigators from the START Study sent this letter in response to the email I distributed yesterday (which follows their letter below) with the headline which I had put in the email Subject "Is Start Study A Waste of Resources?". I am pleased to receive their letter and forward it to the large and broad global NATAP listserve readership, and I will post it on the NATAP website; because I believe in debate and discussion particularly on important issues like "When To Begin HAART" and the START Study. A number of years ago there was a proposal to conduct a "When To Begin Study" and the ACTG and DAIDS held a one day meeting to discuss the pros and cons of such a study. Ultimately it was decided not to conduct that study for many reasons one of which was the thinking that the study might not be able to answer the question and the ART drugs used at the time would be different several years later probably even before the study was completed so the study results may not be applicable. The conditions today are different in that we have many new drugs that are easier to tolerate and safer and more convenient. At this time we do not have a bunch of new ART drugs on the horizon for development and a recent development is that in the past year or so we have been reporting from studies that non-AIDS events are emerging and the development of non-AIDS events appear tied with CD4 count, inflammation, and viral load. High CD4 count and undetectable viral load appear protective from inflammation and developing non-AIDS events such as cardiovascular disease, hepatitis perhaps, kidney disease developed in the interruption arm in SMART, and other comorobidities may be associated with CD4 and viral load such as neurocognitive/brain/CNS complications; lipodystrophy appears associated with starting HAART with a lower CD4. I don't want to say much more but I only want to end by saying that the purpose of my saying is START A Waste of Resources? is because I am not convinced that we will answer the primary question with START. Healthy debate is useful. Is enrolling START very likely, there are some doubts. Will START be able to prove that starting HAART with 600 CD4s is better than 400 or 350? Will the START results be useful to implementing earlier use of HAART? I am not convinced. However, I don't feel as strong about this as I did when I was strongly against conducting this type of study initially a number of years ago. I think conducting the study today is more likely to be successful than previously. But I have doubts we can answer the question; plus, there have been so many cohort studies that I find convincing enough that HAART ought to be considered to be started earlier. I am fairly well convinced that starting HAART earlier, the earlier the better [even at 800 CD4s] will preserve immune function/D4 count and suppress HIV replication, and protect against comorbidities affecting the brain as well as others. In the end it should be a patient-doctor decision anyway because some of the same reasons not to start earlier persist: the risk of HIV drug resistance; body changes are likely to still be a risk even though a lower risk when starting HAART at a higher CD$ count; non-adherence; changes quality of life once you start therapy, etc. Also, what are the long-term risks of taking HAART over 10 to 30 years, which I don't think START will answer because I don't think START will go on for 20 years duration.
Jules Levin.
Here is the letter from the START investigators.
December 18th 2008
Dear Jules,
We saw your recent editorialized comment asking the rhetorical question of whether the "Strategic Timing of Antiretroviral Treatment (START)" trial is a waste of money?. We - the group leading the SMART and START studies - are truly perplexed to see this spin being made to our own research. The article you are referring to and making bolded statements from is a subgroup analysis of the SMART study [1]. The patients included in this analysis were those not on ART at time of entry to SMART, and the analyses found that more serious non-AIDS events occurred in the group randomised to only initiate ART when reaching a CD4 count of less than 250 cells/μL versus those that were randomised to initiate ART immediately (> 350). There are however several important caveats to this analysis that we laid out in the article and these are important to re-emphasise in order to balance your comments. They include: (1) the number of persons included in the analysis were a mere 477 of which 14 developed a serious non-AIDS event - hence the difference in risk of serious non-AIDS events could be a chance finding; only appropriately powered studies such as START can confirm or refute this observation; (2) this was a post-hoc analysis and hence intended to be hypothesis generating and not conclusive - hence it serves as a strong rationale for why START should be done and not the other way around; (3) the CD4 count at which ART was to be re-initiated in the DC arm (deferred ART) of SMART was less than 250 cells/μL whereas in START this count will be at least 100 cells/μL higher. Further, in the VS arm (immediate ART) the majority of patients initiated ART < 500 cells/μl while in START all will initiate > 500 cells/μL.
It has been argued that the money spent on a study like START could be better used on small pathogenesis studies. Actually, carefully designed, appropriately powered, nested case-control studies within cohorts like START also can provide valuable information on disease pathogenesis. Such studies, which can utilize stored specimens collected at study entry and throughout follow-up for participants and link them with clinical events have proven to be critically important in many disease areas. These studies allow comparisons between individuals who did and did not develop the event before the event occurred. SMART has already illustrated the importance of evaluating biomarkers in this way, and the START repository of specimens, including DNA, will undoubtedly provide the basis for important pathogenesis studies of different serious clinical conditions.
In your comments, you quote published editorials for what they think cohort studies have found. Several of us have spend most of our professional carriers performing analyses of observational studies, and in our opinion, the bottom line is that large observational studies have supported findings from randomised controlled trials suggesting that ART protects against AIDS-related opportunistic infections and malignancies, and that this benefit outweighs any risk associated with using ART in untreated patients with CD4 counts less than 300-350 cells/μL; in persons in this situation, the risk of contracting AIDS-related morbidity and mortality increases exponentially with lower CD4 count levels; hence the CD4 count at which ART should be initiated in the deferred arm of START is a CD4 count of 350 cells/μL. Conversely, there is very little direct evidence from randomised controlled trials to support that early initiation of ART protects against non-HIV-related conditions the best evidence comes from the subgroup analysis from SMART discussed above [1]. As non-HIV-related conditions predominate in untreated patients with CD4 count above 350 cells/μL, our opinion is that the current evidence level does not support that the benefit:risk ratio of using ART in treatment nave persons with CD4 count levels above 350 cells/μL (except for specific situations) is favourable. Our interpretation is that it remains uncertain. This uncertainty is however addressable and,will be studied in START which is evaluating the risks and benefits of early treatment with a range of important outcomes, including serious non-HIV-related conditions, AIDS, and health care utilization for an appropriate cost-effectiveness analysis (i.e., one based on a reliable estimate of effectiveness). Once completed, START will inform guidelines of whether and possibly in which subgroups - ART should be initiated in persons who are still in the early stages of HIV infection. If START finds a favourable benefit:risk ratio for early use of ART, then this should increase the impetus for regular testing and increase the proportion of patients on ART. These two factors are likely to have benefits in terms of reducing onward transmission. Population-based approaches to use ART in all diagnosed patients for prevention purposes have been proposed. Without a robust demonstration that the benefits of ART outweigh the risks, regardless of the CD4 count at which it is initiated, it is hard to see how such approaches can justifiably be implemented. Therefore there is an urgent need to complete START as planned.
Thanks for informing the readership of your e-mails of our views to your question.
Jens D. Lundgren, MD
Rigshospitalet & University of Copenhagen, Denmark
Andrew N Phillips, PhD
Royal Free Campus, UCL Medical School, London, UK.
Sean Emery, PhD
University of New South Wales, Sydney, Australia
Fred Gordin, MD
George Washington University, Washington DC, USA
and
James Neaton, PhD
University of Minnesota, Minnesota, USA
1. Strategies for Management of Antiretroviral Therapy (SMART) Study Group. Major clinical outcomes in antiretroviral therapy (ART)-naive participants and in those not receiving ART at baseline in the SMART study. J Infect Dis. 2008 Apr 15;197(8):1133-44.
EDITORIAL COMMENTARY
The Search for Data on When to Start Treatment for HIV Infection
The Journal of Infectious Diseases April 2008;197:1084-1086
Michael D. Hughes and Heather R. Ribaudo
Center for Biostatistics in AIDS Research, Harvard School of Public Health, Boston, Massachusetts
"Data from cohort studies indicate that the risk of AIDS decreases with increasing CD4+ cell counts but nevertheless persists at CD4+ cell counts 500 cells/μL [4]. Furthermore, at any given CD4+ cell count, the risk of AIDS appears to be lower in patients treated with ART than in those who are not [5], and the risk of AIDS or death declines with the start of therapy, even among those with high CD4+ cell counts at initiation, compared with that in patients not receiving ART [6, 7]. Importantly, cohort studies also indicate that risks of serious non-AIDS-related diseases, including cardiovascular, renal, and hepatic conditions and non-AIDS malignancies, are less common at higher CD4+ cell counts [8-13]. Data from the SMART study are consistent with these observations insofar as the rates of AIDS and serious non-AIDS events were significantly higher in participants randomized to the intermittent ART group than in those randomized to the continuous ART group [14]."
"from SMART sub-study in 'naives' below: "Our data suggest that early ART use may result in substantial reductions in serious non-AIDS events.....The risk of a composite outcome was >4-fold higher in DC participants than in VS participants. The data indicate that the risk of such events is reduced by early versus deferred use of ART (the absolute risk of an OD or serious non-AIDS event in DC versus VS participants was 7.0 vs. 1.6/100 person-years).....Our data suggest that early ART use may result in substantial reductions in serious non-AIDS events. the data suggest that reduced risk for serious non-AIDS outcomes arising from the early use of ART might substantially outweigh any increased risk of serious non-AIDS outcomes associated with the use of ART. We believe the data presented here provide a strong clinical rationale for the design and conduct of a randomized trial to estimate precisely the risks and benefits of the initiation of ART before the CD4+ cell count declines to <350 cells/μL for the treatment of chronic HIV disease."
"What is uncertain is whether such a trial is feasible...... With current treatment guidelines moving more toward the recommendation that ART be initiated in subjects with CD4+ cell counts <350 cells/μL rather than waiting for counts <200 cells/μL [12], such a trial would need to enroll asymptomatic subjects with a CD4+ cell count >500 cells/μL to provide reasonable separation between the immediate ART and deferred ART strategies. In a recent study, the median time from presenting with a count between 500 and 650 cells/μL to having a count <350 cells/μL was 2.5 years, with 75% and 90% <350 cells/μL within 5 and 8 years, respectively [13]. Thus, a trial would need to follow all subjects for at least 5 years to quantify just the initial benefits, if any, of immediate ART over deferred ART. Such a duration may also allow the assessment of whether any benefit is transient (as was seen in comparisons of immediate vs. deferred zidovudine monotherapy......given the cost of ART and associated care, it is unlikely to be adequate to show only that earlier initiation of ART is superior to deferred initiation with respect to one or more such end points; rather, it will be important to establish that any difference is sufficiently large (e.g., greater than a 20% reduction in serious morbidity) and is sustained (rather than transient), to justify the costs. In addition to raising important issues for the informed consent process, this necessarily increases the size of the trial needed-a study population in excess of 10,000 participants might be required"
The question of when to start antiretroviral therapy (ART) for infection with HIV has been much discussed, and treatment management guidelines have oscillated from the "hit hard, hit early" philosophy to conservative approaches of deferring treatment in asymptomatic patients until CD4+ cell counts are <200 cells/μL [1, 2], despite their being few data on the subject [3]. In this issue of the Journal, results from a subgroup analysis [4] of a randomized trial, the Strategies for Management of Antiretroviral Therapy (SMART) study, are presented that provide some insight into this question.
Major reasons why, in recent years, guidelines have tended to be more conservative and have recommended deferring treatment until CD4+ cell counts are <200 cells/μL include the following [5]: the low absolute risk of AIDS-defining clinical events at higher CD4+ cell counts; the negative impact of side effects on quality of life as well as the occasional occurrence of life-threatening adverse effects, including ones that have been associated with long-term use of ART (such as myocardial infarction); and the inconvenience of drug regimens, leading to reduced adherence, an increased risk of drug resistance, and limitations on future drug options. Revisiting these issues, Phillips et al. [5] recently concluded that advances in drug development and an improved understanding of the durability and adverse effects of specific drugs has led to renewed interest in the possibility that initiation of ART at CD4+ cell counts >200 cells/μL might be preferable.
The SMART study [6] enrolled HIV-infected subjects with CD4+ cell counts >350 cells/μL (including subjects with prior AIDS events) and randomized participants to a continuous ART strategy versus a CD4+ cell count-guided strategy for interrupting therapy during periods of higher CD4+ cell counts. The study population was predominantly and extensively ART experienced. In this issue, results are presented for a subgroup of SMART participants not receiving ART at the time of enrollment [4]. Specifically, participants who were ART naive (n=249) or who had not received ART for at least 6 months (n=228) were included. Per the SMART study design, the subjects in the continuous ART arm were in effect randomized to (re)initiate ART immediately, and those in the CD4+ cell count-guided arm were randomized to defer (re)initiation until their CD4+ cell count fell to <250 cells/μL. Follow-up was for a mean of 18 months. In this subgroup, as in the overall SMART population, subjects who immediately (re)initiated ART experienced fewer grade 4 symptomatic adverse events than did those deferring (re)initiation. In addition, also as in the overall study population, fewer subjects randomized to (re)initiate ART immediately experienced opportunistic disease (OD) events or died (5 vs. 15 participants; p=.02) or experienced major cardiovascular, renal, or hepatic events (the so-called serious non-AIDS events; 2 vs. 12 participants; p=.01). Of importance, this comparison was driven primarily by events among subjects who had previously received ART (for a median of 4 years). The data that more truly address the question of when first to start ART were limited. Of the 249 ART-naive participants, 1 versus 4 participants experienced OD events or died in the immediate versus deferred ART arms, and 1 versus 4 experienced serious non-AIDS events.
That use of ART at higher CD4+ cell counts might reduce the incidence of OD is not surprising: randomized trials conducted in the early 1990s showed a benefit [7], and observational studies have suggested that this is true for regimens in use more recently [8]. What is more important is that this subgroup analysis raises the possibility that earlier treatment has a beneficial effect, at least in the short-term, not only with respect to OD but also with respect to other serious morbidities that might be associated with HIV infection, without the effect being counterbalanced by severe toxicity.
It is concluded by the authors of the article that the findings of the subgroup analysis need to be validated in a large, randomized clinical trial. We agree-although, to be specific, such a trial should restrict enrollment to subjects who have no history of ART, because this is the population of real interest for the when-to-start question. Subjects who have previously taken ART, even if they have not done so for several months at the time of enrollment, likely have a distinctly different risk profile for AIDS and possibly other events (as seen in the subgroup analysis) and, hence, drive the results of any trial (as also seen in that analysis). Furthermore, there is the potential for differing effects of ART among subjects who have previously received ART versus those who have never received ART-due, for example, to the possibility of the reemergence of archived resistant virus [9-11].
Such trials are under way. For example, one that involves a collaboration between the HIV Prevention Trials Network and the AIDS Clinical Trials Group (HPTN 052/ACTG 5245) highlights another potential benefit of earlier treatment: reducing the risk for transmission. The trial is being undertaken in countries with limited resources, with 50% of the enrollment expected to come from sub-Saharan Africa. Serodiscordant couples for whom the HIV-positive index subject does not meet World Health Organization guidelines for initiating ART and has a CD4+ cell count between 350 and 550 cells/μL are enrolled. The index subject is randomized to start ART immediately or to defer initiation until either his or her CD4+ cell count falls to <250 cells/μL or he or she experiences an AIDS-defining illness. Although originally designed to evaluate the effects of ART on HIV transmission, by following the HIV-positive index cases for clinical outcomes (over at least 5 years) the study will also provide important data to address the when-to-start question.
Along with many other factors, differences in HIV-related morbidity and comorbidities across different regions of the world mean that there is a need for a randomized trial in countries with more-extensive resources. What is uncertain is whether such a trial is feasible. With current treatment guidelines moving more toward the recommendation that ART be initiated in subjects with CD4+ cell counts <350 cells/μL rather than waiting for counts <200 cells/μL [12], such a trial would need to enroll asymptomatic subjects with a CD4+ cell count >500 cells/μL to provide reasonable separation between the immediate ART and deferred ART strategies. In a recent study, the median time from presenting with a count between 500 and 650 cells/μL to having a count <350 cells/μL was 2.5 years, with 75% and 90% <350 cells/μL within 5 and 8 years, respectively [13]. Thus, a trial would need to follow all subjects for at least 5 years to quantify just the initial benefits, if any, of immediate ART over deferred ART. Such a duration may also allow the assessment of whether any benefit is transient (as was seen in comparisons of immediate vs. deferred zidovudine monotherapy [7], despite longer-term sustained differences in CD4+ cell count [14, 15]). However, given the potency of many of the currently available ART regimens, very few subjects randomized to an immediate ART strategy would be expected to exhaust treatment options over this time frame. Thus, such a trial could not evaluate the long-term downsides of immediate therapy (and, hence, the benefits of deferred ART) that are due either to toxicity from cumulative drug exposure or to accumulated viral resistance and exhaustion of treatment options leading to an increased risk of HIV-related morbidity. It is likely unrealistic to undertake a randomized trial that could evaluate such very long-term outcomes. It is also unlikely that observational studies will generate answers to such long-term questions without considerable doubt arising with regard to confounding factors. Realistically, such very long-term outcomes will need to be evaluated by the type of computer-simulation models used in cost-effectiveness analysis [16, 17]-although the validity of such long-term models cannot be verified empirically.
The SMART study was important because it showed that the serious morbidity that was considered a priori to be related to ART might, in fact, be beneficially affected by ART, perhaps because such morbidity is a consequence of chronic HIV infection or because some adverse effects of ART might be less frequent when ART is initiated at higher CD4+ cell counts [18]. Because a major argument against the initiation of ART at higher CD4+ cell counts has concerned the risk-benefit ratio for ART when the notion of benefit tended to be focused on AIDS-defining events and mortality, results from the SMART study emphasize the need to look at serious morbidity more broadly in a when-to-start trial, either as a set of coprimary end points or as a composite end point. Whichever, given the cost of ART and associated care, it is unlikely to be adequate to show only that earlier initiation of ART is superior to deferred initiation with respect to one or more such end points; rather, it will be important to establish that any difference is sufficiently large (e.g., greater than a 20% reduction in serious morbidity) and is sustained (rather than transient), to justify the costs. In addition to raising important issues for the informed consent process, this necessarily increases the size of the trial needed-a study population in excess of 10,000 participants might be required. A major hurdle will be finding such a number of participants who, along with their physicians, are willing to be randomized either to commence ART immediately or to defer its use. To succeed, a when-to-start trial will therefore likely need to involve international collaboration and to use a design that facilitates enrollment, perhaps of very small numbers of subjects at numerous locations in many countries, with simple follow-up and high retention.
Major Clinical Outcomes in Antiretroviral Therapy (ART)-Naive Participants and in Those Not Receiving ART at Baseline in the SMART Study
The Strategies for Management of Antiretroviral Therapy (SMART) Study Groupa
ABSTRACT
Background. The SMART study randomized 5472 human immunodeficiency virus (HIV)-infected patients with CD4+ cell counts >350 cells/μL to intermittent antiretroviral therapy (ART; the drug conservation [DC] group) versus continuous ART (the viral supression [VS] group). In the DC group, participants started ART when the CD4+ cell count was <250 cells/μL. Clinical outcomes in participants not receiving ART at entry inform the early use of ART.
Methods. Patients who were either ART naive ( ) or who had not been receiving ART for 6 months ( ) were analyzed. The following clinical outcomes were assessed: (i) opportunistic disease (OD) or death from any cause (OD/death); (ii) OD (fatal or nonfatal); (iii) serious non-AIDS events (cardiovascular, renal, and hepatic disease plus non-AIDS-defining cancers) and non-OD deaths; and (iv) the composite of outcomes (ii) and (iii).
Results. A total of 477 participants (228 in the DC group and 249 in the VS group) were followed (mean, 18 months). For outcome (iv), 21 and 6 events occurred in the DC (7 in ART-naive participants and 14 in those who had not received ART for >6 months) and VS (2 in ART-naive participants and 4 in those who had not received ART for >6 months) groups, respectively. Hazard ratios for DC vs. VS by outcome category were as follows: outcome (i), 3.47 (95% confidence interval [CI], 1.26-9.56; p=.02); outcome (ii), 3.26 (95% CI, 1.04-10.25; p=.04); outcome (iii), 7.02 (95% CI, 1.57-31.38; ); and outcome (iv), 4.19 (95% CI, 1.69-10.39; p=.002).
Conclusions. Initiation of ART at CD4+ cell counts >350 cells/μL compared with <250 cells/μL may reduce both OD and serious non-AIDS events. These findings require validation in a large, randomized clinical trial.
The most recent randomized clinical trial that addressed the question of when to start antiretroviral therapy (ART) for the treatment of HIV disease was the Concorde study [1]. Since then, despite enormous advances afforded through the development of antiretroviral drugs and the use of combination regimens, there has been no randomized study to address this critically important issue. The absence of randomized data in both the developed and developing world has required antiretroviral treatment guidelines to rely on the interpretation of nonrandomized studies and expert opinions [2, 3]. Current recommendations for the initiation of ART are conservative because of the uncertainty about the risk-benefit ratio associated with earlier ART use at higher CD4+ cell counts. These include toxicities, cost-effectiveness, quality of life issues, adherence, and drug resistance. At present, the initiation of ART is recommended for asymptomatic patients with CD4+ cell counts between 201 and 350 cells/μL or, in developing/resource-poor countries, when the CD4+ cell count has declined to <200 cells/μL.
Data from cohort studies indicate that the risk of AIDS decreases with increasing CD4+ cell counts but nevertheless persists at CD4+ cell counts >500 cells/μL [4]. Furthermore, at any given CD4+ cell count, the risk of AIDS appears to be lower in patients treated with ART than in those who are not [5], and the risk of AIDS or death declines with the start of therapy, even among those with high CD4+ cell counts at initiation, compared with that in patients not receiving ART [6, 7]. Importantly, cohort studies also indicate that risks of serious non-AIDS-related diseases, including cardiovascular, renal, and hepatic conditions and non-AIDS malignancies, are less common at higher CD4+ cell counts [8-13]. Data from the SMART study are consistent with these observations insofar as the rates of AIDS and serious non-AIDS events were significantly higher in participants randomized to the intermittent ART group than in those randomized to the continuous ART group [14].
Here, we report the major clinical outcomes for a subgroup of participants enrolled in SMART who were either ART naive or who had not been receiving ART for 6 months before randomization. This subgroup provides a randomized comparison between early ART at a CD4+ cell count of >350 cells/μL versus deferred ART until the CD4+ count declines to <250 cells/μL. Subgroup analyses for major clinical outcome according to ART history were proposed a priori. However, the analyses presented here are post hoc. Thus, we regard our results as hypothesis generating.
Results
Patient population. A total of 477 participants (8.7%) in the SMART study were either ART naive ( m=249) or had not been receiving ART for >6 months before randomization (n=228) and are included in these analyses. The relationship of these participants to the total SMART cohort is summarized in figure 1.
Table 1. Overall baseline characteristics and those for patients who were antiretroviral therapy (ART) naive or who had not received ART for >6 months (no ART) before randomization in the SMART study.
The median duration of ART exposure in previously treated participants was 4 years. Approximately 75% had used an HIV protease inhibitor (PI), for a median duration of 2.5 years. Approximately 50% had used a nonnucleoside reverse-transcriptase inhibitor (NNRTI), for a median duration of 2.5 years. All previously treated participants had used a nucleoside reverse-transcriptase inhibitor (NRTI).
In the VS group, (re)initiation of ART comprising NRTIs in combination with NNRTIs, NRTIs in combination with a PI, and NRTIs alone was reported for 134 (56%), 73 (31%), and 31 (13%) participants, respectively.
There were modest differences in baseline characteristics between participants who were ART naive and those who were experienced. ART-naive participants were on average 3 years younger, less likely to be current smokers, less likely to be taking blood-pressure-lowering or lipid-lowering therapy, and had slightly lower serum triglyceride levels. They were also less likely to be coinfected with either hepatitis B or C virus and less likely to have had a prior AIDS-defining illness.
The mean period of follow-up was 18 months (median, 15 months). Eleven (4 DC and 7 VS) participants were lost to follow-up.
Exposure to ART. DC and VS participants received ART for 18% and 89% of the follow-up time, respectively. In the DC group, 68 participants (30%) (re)initiated ART, 25 of whom were ART naive (21% of the ART-naive participants) and 43 of whom were ART experienced (39% of the ART-experienced participants). Reasons for (re)initiation of ART included CD4+ cell count decline to <250 cells/μL (38 participants [56% of those who (re)initiated ART]); CD4+ cell percentage decline to <15% (28 participants [41%]); symptoms of HIV disease or development of OD (14 participants [21%]), with 4 participants developing an OD event; and nonprotocol reasons (14 participants [21%]). Of these 68 participants, 30 (44%) initiated ART at a CD4+ cell count >250 cells/μL. Kaplan-Meier estimates for DC participants (re)initiating ART were 11%, 25%, 39%, 45%, 47%, and 55% at 6, 12, 18, 24, 30, and 36 months, respectively. In the DC group, participants (re)initiated ART at a median CD4+ cell count of 236 cells/μL (interquartile range [IQR], 183-310 cells/μL); those who were ART naive initiated ART at median of 245 cells/μL (IQR, 192-338 cells/μL), versus 225 cells/μL (IQR, 180-307 cells/μL) among ART-experienced participants. In comparison, 238 participants (96%) in the VS group (re)initiated ART at study entry, at a CD4+ cell count 200 cells/μL higher; 189 (79%) had a CD4+ cell count between 350 and 549 cells/μL, and 49 (21%) had a CD4+ cell count >550 cells/μL.
In the DC group, 68 (30%) of 228 participants started ART, and 27 (12%) subsequently interrupted ART. The majority of follow-up time was accrued before the first (re)initiation of ART: 236 person-years (79%) accrued from study entry to first (re)initiation of ART, 39 person-years (13%) from (re)initiation of ART to discontinuation, and 25 person-years (8%) after discontinuation of ART. During this 8% of follow-up time, 3 participants in the DC group experienced an OD event, a serious non-AIDS event, or died.
CD4+ cell count and plasma HIV RNA outcomes. Reflecting the differing ART exposures, there were substantial differences in CD4+ cell counts and plasma HIV RNA levels over time between the DC and VS groups. Figure 2 shows time curves for CD4+ cell count and plasma HIV RNA level for each treatment group. The CD4+ cell count was, on average, 148 cells/μL lower in DC than VS participants. The proportions of participants with plasma HIV RNA levels <400 copies/mL at 12 and 24 months were 10% and 25% in the DC group and 59% and 56% in the VS group.
Figure 2. Mean change from baseline CD4+ cell count (A) and percentage of patients with plasma HIV RNA levels <400 copies/mL (B) during follow-up. The solid red lines indicate participants in the drug conservation (DC) group, and the dashed blue lines indicate participants in the viral supression (VS) group. Vertical bars represent ± 2 standard errors.
Clinical outcomes. Fifteen participants in the DC group and 5 in the VS group experienced OD/death (HR, 3.5 [95% confidence interval {CI}, 1.3-9.6) (figures 3 and 4A). The corresponding HR(DC/VS) for OD and for serious non-AIDS events was 3.3 (95% CI, 1.0-10.3) and 7.0 (95% CI, 1.6-31.4), respectively (figures 3, 4B, and 4C). There was relatively little overlap in OD and serious non-AIDS events (table 2); 21 participants in the DC group and 6 in the VS group experienced at least 1 of these (the composite end point) (figures 3 and 4D). The HR for the composite outcome was 4.2 (95% CI, 1.7-10.4). The excess risk (DC minus VS) was 5.4 events/100 person-years (95% CI, 3.4-7.4). The cumulative percentage developing the composite end point by 12, 24, and 36 months were 8%, 13%, and 18% for DC participants and 2%, 4%, and 4% for VS participants (figure 4D).
Figure 3. Hazard ratios (HRs) for clinical outcomes among the overall cohort and among those who were either antiretroviral therapy (ART) naive or not receiving ART at baseline. Horizontal lines represent the 95% confidence interval for each HR. The P value for interaction compares event rates within each category for ART-naive and ART-experienced participants. DC, drug conservation; VS, viral supression.
For the composite end point, HRs were 4.6 (95% CI, 1.0-22.2) for the ART-naive participants and 3.7 (95% CI, 1.2-11.2) for the ART-experienced participants ( for difference in these HRs). In participants excluded from this cohort (see figure 1), HRs for the composite end point were 1.3 (95% CI, 0.6-3.2) for experienced participants not receiving ART at entry (200 DC and 199 VS participants) and 2.0 (95% CI, 1.5-2.6) for participants receiving ART at entry (2292 DC and 2304 VS participants) ( for the difference in HRs between these 4 subgroups). In the entire SMART study, there were 270 participants with a composite end point (180 DC and 90 VS participants), with a corresponding HR of 2.1 (95% CI, 1.6-2.7).
Table 3 summarizes the composite outcome by treatment group and proximal CD4+ cell count. Event rates were higher during the time when participants experienced lower CD4+ cell counts. Per design, participants in the DC group, relative to those in the VS group, spent substantially longer periods of time with CD4+ cell counts in the <350 cells/μL stratum (28% vs. 11% of the time) and substantially less time with CD4+ cell counts in the 500 cells/μL stratum (32% vs. 61% of the time). The majority of events in the DC group (57%) occurred at proximal CD4+ cell counts 350 cells/μL. Nine (75%) of 12 events in the DC group (1 OD and 8 serious non-AIDS events) that occurred at CD4+ cell counts <350 cells/μL happened before the (re)initiation of ART. Table 3 also summarizes the composite outcome in the subgroup of participants without prior AIDS and without a prior non-AIDS-defining illness, for which the HR(DC/VS) was 6.0 (95% CI, 1.7-20.8) and 7.5 (95% CI, 2.2-25.4), respectively. Although event rates were similar or reduced compared with the overall cohort, the HR(DC/VS) did not decrease after excluding participants with these prior events.
For the composite end point, HRs were 4.6 (95% CI, 1.0-22.2) for the ART-naive participants and 3.7 (95% CI, 1.2-11.2) for the ART-experienced participants (p-.84 for difference in these HRs). In participants excluded from this cohort (see figure 1), HRs for the composite end point were 1.3 (95% CI, 0.6-3.2) for experienced participants not receiving ART at entry (200 DC and 199 VS participants) and 2.0 (95% CI, 1.5-2.6) for participants receiving ART at entry (2292 DC and 2304 VS participants) (p=.27 for the difference in HRs between these 4 subgroups). In the entire SMART study, there were 270 participants with a composite end point (180 DC and 90 VS participants), with a corresponding HR of 2.1 (95% CI, 1.6-2.7).
Table 3 summarizes the composite outcome by treatment group and proximal CD4+ cell count. Event rates were higher during the time when participants experienced lower CD4+ cell counts. Per design, participants in the DC group, relative to those in the VS group, spent substantially longer periods of time with CD4+ cell counts in the <350 cells/μL stratum (28% vs. 11% of the time) and substantially less time with CD4+ cell counts in the >500 cells/μL stratum (32% vs. 61% of the time). The majority of events in the DC group (57%) occurred at proximal CD4+ cell counts >350 cells/μL. Nine (75%) of 12 events in the DC group (1 OD and 8 serious non-AIDS events) that occurred at CD4+ cell counts <350 cells/μL happened before the (re)initiation of ART. Table 3 also summarizes the composite outcome in the subgroup of participants without prior AIDS and without a prior non-AIDS-defining illness, for which the HR(DC/VS) was 6.0 (95% CI, 1.7-20.8) and 7.5 (95% CI, 2.2-25.4), respectively. Although event rates were similar or reduced compared with the overall cohort, the HR(DC/VS) did not decrease after excluding participants with these prior events.
We investigated whether differences in nadir CD4+ cell counts might have affected the distribution of clinical events. The data in table 1 indicate that median nadir CD4+ cell counts were not substantially different for the ART-naive or previously treated participants between treatment groups. The differences between ART-experienced and ART-naive participants with respect to median baseline CD4+ cell count (437 and 376 cells/μL, respectively) and median nadir CD4+ cell count (452 and 341 cells/μL, respectively) participants were not appreciable. The HR(DC/VS) for our primary outcome was unaffected in analyses that did or did not adjust for nadir CD4+ cell count (HR of 3.7 in both analyses).
Grade 4 events were experienced by 25 DC participants and 18 VS participants. The HR(DC/VS) was 1.6 (95% CI, 0.9-3.0).
Discussion
The randomization in the SMART study for this patient subset directly compared the early use of ART (at >350 CD4+ cells/μL) versus the deferred use of ART (at <250 CD4+ cells/μL). To our knowledge, the data presented here represent the first substantive evidence from a randomized trial to inform this critical issue in HIV medicine since 1994 [1]. During follow-up, there was substantially lower exposure to ART in the deferred treatment group (DC). The deferred use of ART was consistent with the protocol in the majority of participants; although the DC strategy included discontinuation of ART when the CD4+ cell count increased to >350 cells/μL, 92% of accrued follow-up was in participants who were either not receiving ART or taking ART for the first time after randomization. In the early ART group (VS), most participants (re)initiated ART with CD4+ cell counts between 350 and 549 cells/μL. We combined data from participants who were ART naive and those who had not been receiving ART for at least 6 months before randomization. Had we been more conservative and included only participants who had not been receiving ART for >12 months before randomization, we would have reduced the overall number of composite events from 27 (21 in DC and 6 in VS participants) to 23 (17 in DC and 6 in VS participants). The calculated HR(DC/VS) would still have indicated a significant benefit for the VS strategy (HR, 3.1 [95% CI, 1.2- 7.9]).
The results suggest that morbidity and mortality associated with HIV is probably higher than previously thought for patients with earlier stages of HIV disease. In DC participants, the rates of OD and serious non-AIDS events were similar. The risk of a composite outcome was >4-fold higher in DC participants than in VS participants. The data indicate that the risk of such events is reduced by early versus deferred use of ART (the absolute risk of an OD or serious non-AIDS event in DC versus VS participants was 7.0 vs. 1.6/100 person-years). This difference in absolute risk, if confirmed in a larger trial, is important. Treatment guidelines for the initiation of ART are conservative because the risk of AIDS (OD in our categories) at high CD4+ cell counts is relatively low and because ART-related toxicities (and other factors) have been assumed to outweigh any likely ART-related benefits. Our data suggest that early ART use may result in substantial reductions in serious non-AIDS events. Relative to the DC group, 289 additional person-years of ART prevented 20 serious clinical events in the VS group. Hence, prevention of 1 event requires 14 extra person-years of ART. When assessed for AIDS events alone, prevention of 1 event requires 42 extra person-years of ART.
The occurrence of AIDS at higher than expected CD4+ cell counts has been described previously [7, 17]. There are fewer data linking CD4+ cell counts and the risk of non-AIDS-related morbidity and mortality. Some non-AIDS conditions appear to be linked to progressive immunodeficiency [8, 9, 18]. In the SMART study, lower CD4+ cell counts and higher plasma HIV-RNA levels explain much of the excess risk in DC compared with VS participants for non-OD death as well as OD [18]. The data presented here for SMART participants not receiving ART are consistent with these observations.
Mechanistically, it appears to be unlikely that a common pathway for all non-AIDS events exists. Furthermore, the risk and/or the benefit of ART for reducing non-AIDS diseases is not clear. For example, PI therapy has been associated with an increased risk of myocardial infarction [19, 20]. On the other hand, continuous ART in SMART was associated with better cardiovascular outcomes than ART interruption [14]. It is possible that untreated HIV disease supports a range of inflammatory responses that, particularly over long periods, set the scene for atherosclerosis in individuals with additional risk factors for poor cardiovascular outcomes. The increased risk of cardiovascular disease in patients with systemic lupus erythematosus is similar in this regard [21, 22]. In addition, in SMART, changes in lipids appear to explain some of the increased risk of cardiovascular disease in the DC compared with the VS group [23]. The pathogenesis of hepatic and renal disease along with non-AIDS cancer is less clear. These equivocal and perhaps controversial issues further highlight the need for a careful evaluation of risks and benefits of early treatment with ART [24]. Furthermore, they reinforce the importance of studies with systematically collected clinical end points matched with appropriate biological samples as a key feature of trial design for HIV infection [25].
There are limitations to our findings. First, we did not prespecify the patient subgroup or composite outcome of OD and serious non-AIDS events in the original SMART analysis plan. Thus, our findings generate new hypotheses that require confirmation. Predefined components of our composite outcome have been reported for all SMART participants [14, 26]. For each end point, there is an excess risk in the DC compared with the VS group. The higher risk in DC compared with VS participants for this composite outcome is not limited to those in this small subgroup. The HR(DC/VS) for the composite of OD and serious non-AIDS events is >1 for each of the 4 cohorts classified by baseline ART use and is and significantly >1 for all SMART participants. Second, although the CIs around the HR estimates are wide as a result of the small number of composite events (n=27), they are qualitatively similar to the reliably estimated 2-fold difference in risk for the overall SMART study. Third, we included a small number of participants who had prior AIDS-defining illnesses (23 DC and 29 VS participants). However, analyses that excluded these participants were consistent with the present results (table 3). Finally, in SMART we did not formally collect information on treatment-limiting toxicities that did not attain the severity to qualify as a grade 4 adverse events. During the period of observation, there were 3 DC participants (1%) and 24 VS participants (10%) who ceased ART with toxicities being the reason cited. During follow-up, there were 25 and 18 grade 4 events in the DC and VS groups, respectively.
These data indicate that the absolute risk of serious non-AIDS events in untreated patients is greater than the risk of AIDS at high CD4+ cell counts. They reveal that risk reductions for AIDS associated with the earlier use of ART appear to be greater than previously estimated. Furthermore, the data suggest that reduced risk for serious non-AIDS outcomes arising from the early use of ART might substantially outweigh any increased risk of serious non-AIDS outcomes associated with the use of ART. We believe the data presented here provide a strong clinical rationale for the design and conduct of a randomized trial to estimate precisely the risks and benefits of the initiation of ART before the CD4+ cell count declines to <350 cells/μL for the treatment of chronic HIV disease.
Methods
Participants. Participants were included if they were randomized in the SMART study and were either ART naive or satisfied the following eligibility criteria for having been previously treated but not receiving ART for a minimum of 6 months before randomization (hereafter, "not receiving ART"): (1) were not receiving ART at the time of randomization in SMART; (2) had at least 1 plasma HIV RNA assessment during the 6 months before randomization; and (3) all plasma HIV RNA levels in the 6 months before randomization were >10,000 copies/mL. We excluded patients who had not been receiving ART for <6 months before randomization to avoid the possible short-term effects of ceasing ART (e.g., the steeper decline in CD4+ cell count). The principal eligibility criterion for the SMART study was a CD4+ cell count of >350 cells/μL. Other eligibility criteria for SMART are described elsewhere [14].
Randomization. Participants were randomly allocated in equal proportions to 1 of 2 treatment strategies: continuous ART (the viral supression [VS] group) and intermittent ART (the drug conservation [DC] group). For the present analyses, this randomization corresponded to the immediate (re)initiation of ART after randomization at CD4+ cell counts >350 cells/μL (the VS group) versus the deferred (re)initiation of ART when either the CD4+ cell count declined to <250 cells/μL, the CD4+ cell percentage declined to <15%, or symptoms of HIV disease developed (the DC group).
Assessments. Before randomization, participants' ART and medical history were obtained along with nadir and baseline CD4+ cell counts and plasma HIV RNA levels. Routine visits occurred at 1 and 2 months, every 2 months thereafter for the first year, and every 4 months in subsequent years. Visits included clinical assessments, and samples were obtained for measurement of CD4+ cell count and plasma HIV RNA level. At baseline and annually, a 12-lead electrocardiogram was obtained, and data were electronically transmitted to a central reading facility for assessment of silent myocardial infarction [15].
End-point definitions. Definitions for the primary end point of opportunistic disease (OD) or death from any cause (OD/death) and the major secondary composite end point of cardiovascular-, renal-, and hepatic-disease events are given in the online-only appendix of the primary-results article [14]. In the present report, 4 major clinical outcomes are considered: (i) the SMART primary end point (OD/death); (ii) OD (fatal or nonfatal); (iii) serious non-AIDS events (cardiovascular, renal, and hepatic disease plus non-AIDS-defining cancers [excluding nonmelanoma skin cancers] and deaths from non-OD causes); and (iv) the composite of outcomes (ii) and (iii). In the end-point categories (ii) and (iii), deaths due to OD and all causes except OD are counted. Thus, the composite of (ii) and (iii) can be viewed as major morbidity (OD and non-OD) or all-cause mortality. As previously reported, an end-point review committee, blinded to treatment group, reviewed all major nonfatal clinical events [14]. The end-point review committee also classified the underlying cause of death using the system of the Coding of Death in HIV Project [16]. Grade 4 adverse events were defined as potentially life-threatening symptomatic events according to the Division of AIDS common toxicity table.
Statistical analysis. All analyses were intention to treat. Data through 11 January 2006 were reported. On that date, following the recommendations of an independent data and safety monitoring board, all participants in the DC group were advised to (re)initiate ART. Thus, time-to-event analyses are censored at the earliest of the date of death, the lost to follow-up date, or 11 January 2006.
The time from randomization to first event was compared between the DC and VS groups by use of Kaplan-Meier plots for each event category. For patients reporting multiple events in a single outcome category, only the first event was included. For patients reporting multiple events in different outcome categories, each first event in each category was included. The hazard ratio for DC vs. VS-hereafter, HR(DC/VS)-was estimated by fitting a Cox model with a single binary indicator for treatment group overall and separately in ART-naive and ART-experienced participants. Differences between these 2 groups were assessed by including an interaction term in the Cox model. Randomization protects these comparisons between treatment groups from confounding.
Person-years at specific CD4+ cell counts were counted according to the most recently measured or time-updated value (latest CD4+ cell count), and rates per 100 person-years for the composite outcome were determined for DC and VS participants. Randomization does not protect these comparisons between treatment groups from confounding.
Throughout, P values refer to 2-sided tests. Because the subgroup and the serious non-AIDS outcomes considered in this report were not prespecified, we recommend cautious interpretation of P values.
|
|
| |
| |
|
|
|
